Epi Final Study Guide

Systematic Reviews: Meta-Analyses

- "Study of Studies" - Method used in Cochrane Reviews and ARHQ Reviews - Very powerful methodology - Stringent criteria for inclusion in the analysis - Quantitative analysis to combine and compare results of the studies

Non-randomized clinical trials (quasi-experiments)

- No randomization to treatment groups - Comparison group may be different from the intervention group - Control of the intervention

Which of the following research designs are observational studies?

- Prospective cohort studies - Cross-sectional studies - Retrospective cohort studies

Randomized clinical trials

- Randomization to treatment groups - Control of the intervention - Examples include - Almost all studies of medications - Most studies of devices - Many studies of interventions

Prospective study designs

- Randomized clinical trials - Non-randomized clinical trials (quasi-experiments) - Cohort studies

Which of the following p-values are statistically significant when alpha is set at 0.05? SELECT ALL THAT APPLY

0.01 0.04

Two types of samples

1. Probability 2. Nonprobability

Steps to hypothesis Testing

1. State the null and alternative hypotheses 2. Determine the significance level to be used (alpha) 3. Determine what statistical test to use 4. Calculate the test statistic and determine the p-value 5. Apply your decision rule regarding the level of significance

Prospective Cohort Studies

A group (cohort) of disease-free individuals is identified at one point in time and then followed over a period of time to determine whether whether the outcome (disease) occurs The studies are also called follow-up, longitudinal, or concurrent studies

Hypothesis Testing

A statement is generated about a population and statistics are used to assess the likelihood that the hypothesis is true • Statistical hypotheses • Research hypotheses

Sample

A subset of the population of interest • Used to study a health-state of interest and then use the information to make inferences about the larger population

Calculation of Absolute Risk

Absolute Risk = Number of people who get the disease during a designated period --------------------------------- Population of interest at the start of the designated time period

Research hypothesis

Alternative hypothesis • What the researcher thinks will happen • Called the research or alternative hypothesis • Abbreviated H1 H1: VLBW infants who receive exclusive human milk feedings in the first 28 days of life have a lower risk of developing NEC

Analytic Epidemiology

Analytic epidemiology studies the 'why' or 'how' of health-related events. The purpose of analytic epidemiology is to identify the causes or risk factors for the health-related state or event (good or bad) of interest. Analytic epidemiology studies the causes or risk factors for health-related states or events, also known as the 'determinants' of health and health related events. Analytic epidemiology quantifies the association between exposures and outcomes to determine causal relationships

Which of the following study designs is lowest on the evidence pyramid?

Case Control Study

Which of the following types of research designs is ALWAYS a retrospective study?

Case Control Study

Case Control Study

Case-control study Odds ratio = The odds that a case was exposed ---------------------- The odds that a control was exposed

When performing hypothesis testing, a type II error is made when a researcher incorrectly

Concludes that there is not a significant difference between groups when there really is one.

Which of the following types of study designs provides the lowest level of evidence?

Cross sectional study

Confounding

Defined as "the distortion of a measure of the effect of an exposure on an outcome due to the association of the exposure with other factors that influence the occurrence of the outcome." (Porta, 2008, p. 49) "Confounding occurs when all or part of the apparent association between the exposure and the outcome is in fact accounted for by other variables that affect the outcome and are not themselves affected by the exposure" (Porta, 2008, p. 49)

Descriptive Epidemiology

Descriptive epidemiology is the 'who, where and when' of a health-related event. Descriptive epidemiology describes who is affected by the health-related event, the location of the event, and the timing of the event. Descriptive epidemiology describes the persons, place, and time involved in the health event, also known as the 'distribution' of health and health-related events.

Retrospective Cohort Studies

Have a historical perspective - Requires access to information about the subjects before the exposure to disease causing factors and the onset of disease occurs - Often able to accomplish this when there are comprehensive medical records Also known as historical cohort studies Example of retrospective cohort studies - Fertility drugs and the risk of ovarian cancer studies

Level of Significance

Level of significance or the "critical value" or 'alpha' is usually set at = .05 So, if the calculated p-value is <.05. that means that there is a statistical relationship (e.g., difference, association)

A study of human milk feeding of premature infants finds that the odds ratio for developing sepsis during the infant's hospital course was 0.7 for the infant who received human milk. This finding indicates that

Human milk feeding reduces the risk of sepsis

Statistically Significant Difference

Indicates that the observed difference between samples is large enough to reflect a difference between populations The difference between the two samples from the population is probably not due to chance. If you took two more samples from the same population you should find the difference again Usually alpha is set at 0.05 - Means that there is a probability of less than 5% (<.05) that you find a difference that really does not exist

Inferential Statistics

Inferential statistics enables us to generalize or make inferences from a sample of data to a larger group of subjects or a population

The Bottom of the Pyramid

Opinion and experience

Sample

Is a subset of the population of interest

Alpha

Means that there is a 5% chance (p=.05) that you will incorrectly conclude that there is a difference/association when there is none. Alpha is the probability that you will incorrectly say that there is a difference/association

Measures of Association

Measures of association are used to compare the occurrence of disease in one group with the occurrence of disease in another group. Commonly used measures are - Risk Ratio or Relative Risk Ratio (RR) - Odds Ratio (OR)

convenience sample

Non-probability sample (not for inference). Literally a sample of convenience.

Interpretation of the Odds Ratio

OR = 1 - Odds in the exposed is equal to the odds in the non- exposed OR > 1 - Odds in the exposed group is greater then the odds in the non-exposed group OR < 1 - Odds in the exposed group is less then the odds in the non-exposed group

A ratio of 'odds'

Odds ratio = The odds of an event in one group -------------------------------------------- The odds of the event in a different group

Study Designs in Epidemiology

Systematic Reviews - Meta-analyses Experimental studies - Randomized trials - Non-randomized trials Observational studies - Cohort studies - Case-control studies - Cross-sectional studies

Probability Basics

Probability is a number between 0 and 1.0 Never a negative number Expressed as a percentage P = .01 means that there is a 1% chance P = .05 means that there is a 5% chance

Estimation

Process of determining likely values for unknown population parameter A point estimate for a population parameter is the best single number estimate of that parameter. A confidence interval is a range of values for the estimated population parameter with a level of confidence attached (e.g., 95% confidence that the range or interval contains the parameter).

Sampling Method

Process researchers use to select subjects from the population being studied Never assume that a random sampling method was used

The Framingham Heart Study is an example of what type of research design?

Prospective Cohort Study

Interpretation of the Relative Risk Statistic

RR = 1 - Risk in the exposed is equal to the risk in the non- exposed RR > 1 - Risk in the exposed group is greater then the risk in the non-exposed group RR < 1 - Risk in the exposed group is less then the risk in the non- exposed group

Which of the following types of study designs are considered the "gold standard" for determining the effectiveness of interventions?

Randomized Clinical Trial

Which of the following types of single study designs provides the strongest level of evidence to determine cause and effect?

Randomized Controll Trial

The Women's Health Initiative is an example of what type of research design?

Randomized Controlled Trial

Relative Risk

Relative risk is a measure of association between the exposure to a particular factor and the risk of a particular disease or outcome Relative Risk = Incidence among the exposed -------------------------------- Incidence among the non-exposed

Case-Control Studies

Retrospective study design Studies people who already have the disease and looks backward at exposures Two groups of individuals are studied - A group that has the disease under study (cases) - A group that does not have the disease under study (controls) Their health information is then examined to determine whether there are risk factors that may be related to their current disease

Risk Factors

Risk factors are factors or events associated with the disease or health state or event of interest • Risk factors are not necessarily direct causes of disease or injury but are associated with the development of the disease or injury • Examples of risk factors • Poverty • Neighborhood • Race • Education

Quota Sample

Select a pre-determined number of individuals into sample from groups of interest. Participants are not randomly selected

Which of the sampling method provides the best representation of the population of interest?

Simple random sample

Statistical Hypothesis

Starts with the assumption that there is no difference, or no association between the groups or variables • Called the Null Hypothesis • Abbreviated H0 H0: There is no difference in the incidence of NEC in very low birth weight infants who receive human milk and those who do not receive human milk

Attributable Risk

The amount of disease associated with a causative factor

attributable proportion

The attributable proportion, also known as the attributable risk percent, is a measure of the public health impact of a causative factor. Appropriate use of attributable proportion depends on a single risk factor being responsible for a condition. When multiple risk factors may interact (e.g., physical activity and age or health status), this measure may not be appropriate.

Threats to good care

The care that we provide is inappropriately influenced by - Seduction by Authority - The False Idol of Technology - Letting Sleeping Dogma Lie - The Pursuit of Pedantry - Numerators in Search of Denominators

Level of evidence...

The hierarchy (or pyramid) of evidence ranks study designs - The designs on the top (meta-analyses, randomized trials) provide the strongest or highest levels of evidence - The designs at the bottom (opinion, case report, cross sectional studies) produce the weakest level of evidence.

non probability sampling

The members of the population of interest do NOT have the same opportunity (equal chance) for selection into the study group(s) Used when the researchers cannot use a random sampling method

• Sufficient cause

The presence of the factor or group of factors is enough for the disease to occur • The disease can also occur with other groups of factors

Risk

The probability that an event will occur (Porta, 2008) A measure of disease frequency (Rothman, 2002) • Risk is calculated as a proportion • Risk is presented as a "probability" and can vary from 0 to 1.0 Risk is the terminology used to describe individuals Incidence proportion is the terminology used to describe populations The terms are used interchangeably Risk is the terminology widely used because it it is readily understood and is a familiar concept

When performing a hypothesis test to compare groups, a type I error occurs when

The researcher concludes that there is a significant difference between two groups when there really is not a difference.

Study Deisgn

The study design influences that quality or 'strength' of the evidence produced.

Inferential Statistics

There are two broad areas of statistical inference • Estimation of population parameters - Used to estimate population parameters such as means and proportions • Hypothesis testing - Used to examine data to see if there is... - An "association" between variables - A "difference" between groups

A researcher compares two interventions for preventing surgical site infections on a surgical unit in an academic medical center. A statistical test compring the number of surgical site infections between the two groups yield a p-value of .01. This p-value means

There is a statistically significant difference between the two groups.

Sources of error in estimating sample

There is always random error in any assessment Sample selection • A sample may not accurately represent the population • Random selection from a population is the ideal but there is still always some sampling error Measurement error • Accuracy of the measurements

A new intervention is studied to see if it is assisted with a reduction in pressure ulcers in hospitalized older adults. The relative risk of pressure ulcer was 2.5 in the intervention group. This relative risk statistic means that

There is an increased risk of pressure ulcers in the intervention group.

A researcher compares two interventions for preventing falls on a medical unti in a major academic medical center. A statistical test compring the number of falls between the two groups yield a p-value of .5. This p-value means

There is not a statistically significant difference between the two groups.

DESCRIPTIVE STATISTICS

To describe data

INFERENTIAL STATISTICS

To make inferences

Cochrane reviews are at the top of the evidence pyramid..

True

Randomized controlled trials always evaluate an intervention and include a comparison group

True

Sample size plays an important role in the ability to detect a difference between groups.

True

The Bradford Hill criteria are used to determine disease causation.

True

two types of experimental designs:

randomized clinical trials and quasi-experiments.

Cross-sectional studies

• A study population is assessed at a single point in time • Participants are asked about their current disease state and their exposures to certain factors • Descriptive study • These are often called prevalence studies • Very weak study design!

Casual Pie Components

• Causative agent or event • Characteristics of the person • Characteristics of the environment • Presence of preventative factors • Action of catalysts or additive factors • Numerous mediating and moderating factors

Confidence Interval

• Confidence interval is calculated as the - Point estimate + margin of error • Margin of error has 2 components - Standard error of the estimated parameter - A 'multiplier' - a number based on sample size

Case Reports and Case Series

• Descriptive study design • May generate hypotheses about causation

Sampling Error

• Difference between the sample and the population

Simple Random Sampling

• Enumerate all members of population (sampling frame), select the desired number of individuals at random (each has same probability of being selected)

Stratified Sample

• Organize population into mutually exclusive strata; select individuals at random within each stratum

Populations

• Populations are described using Population Parameters (P for P) • Population parameters are reported using Greek letters (e.g., μ, σ)

Probability Sampling

• Probability sampling = Random sampling • Every member of the population has a chance of being selected • The probability of being selected can be calculated

3 Study Designs in Epidemiology

• Prospective • Retrospective • Cross-sectional

Samples

• Samples are described using Sample Statistics (S for S) • Sample statistics are reported using Roman letters (e.g., M, SD)

Cluster Sample

• Samples clusters or groups instead of individuals • Examples might be schools, clinics, neighborhoods, • Selection process is still random

Important Concepts for Inferential Statistics

• Sampling • Confidence Intervals • Probability • Hypothesis testing

Probability Sampling Methods

• Simple random • Stratified random • Systematic sampling • Cluster sampling

Standard Error

• Standard error (SE or s.e.) is a measure of the accuracy or precision of an estimate of a population parameter (mean or proportion) • Standard error of the mean (SEM) - Describes the accuracy of the sample mean as an estimate of the true population mean • Standard error of a proportion (SE or s.e.) - Describes the accuracy of the sample proportion as an estimate of the true population proportion

Systematic Sample

• Start with sampling frame; determine sampling interval (N/n); select first person at random from first (N/n) and every (N/n) thereafter

Sample Statistics

• Statistics are estimates of the population parameters • When reporting the results of research, always report sample statistics using Roman letters - M instead of μ - SD instead of σ

Bradford Hill Criteria for Causation

• Strength of association • Consistency • Specificity • Temporal relationship • Biological gradient • Plausibility • Coherence • Analogy • Experimental evidence

Causal Mechanisms

• Strength of causal effect • Dose of the exposure • Period of exposure • Interaction between causal factors • Induction time between exposure and outcome • Presence of catalysts and preventatives

Level of evidence

• The 'level of evidence' is determined by the type of the study design used to obtain the evidence • The 'level of evidence' is determined by the type of the study design used to obtain the evidence

• Necessary cause

• The factor must must be present for the disease to occur • The disease does not occur unless the factor is present

Population

• The group of interest... the group that you want to generalize to - Adult men who have type II diabetes - Preterm infants - Older women with osteoporosis

Population Parameters

• You can never know the true population values • You can only estimate parameters • So, DO NOT report parameters!! (e.g., μ, σ)